Data Center is one of the most important parts of information industry. In order to make sure the servers are working in the suitable environment, the refrigeration system in data center must keep running all year round. So it is important to monitor the running status of refrigeration system, and it has significant meaning for the safe operation of data center. This paper presents a fault diagnosis warning system of refrigeration system using Grey Theory and Fault Direction Space Method (FDSM). Considering the particularity of refrigeration system in data center, the Grey Forecast was improved to raise accuracy. And two-phase heat transfer fault direction space method was developed to diagnose the condenser and evaporator faults. With this method, 13 kinds of common faults in refrigeration system can be diagnosed and predicted. A simulation experiment was carried.